AI & SocietyOctober 15, 2026

What the 6 Ds of Exponential Technology Don't Show: Why Industries Facing AI Trust the Wrong Numbers

Radio survived television by giving up the living room. Most industries facing a new technology get a few early years to make a move like that, and most of them spend those years trusting numbers that still look fine.

A woman in a Second World War military-style uniform leans on a glowing Wurlitzer jukebox in a warmly lit dance hall, with a Music Brings Us Together sign and an American flag behind her

During the Second World War, a young American soldier kept noticing the same small thing in restaurants. People fed nickels into the jukebox and picked a song. Then they picked the same song again, and again, though dozens of other records sat in the machine. Most people would have called it bad taste. He saw a business in it.

His name was Todd Storz, and that is how he told the story in 1957. A few years after the war he bought a small radio station in Omaha, one allowed to broadcast only between sunrise and sunset, and built it around that habit: a short list of the songs people played most, repeated all day. Within a few years it was the most popular station in town, and the idea spread across the country under the name it still has, Top 40.

Storz was rebuilding radio at exactly the moment television was taking it apart. For two decades, families had gathered around the radio in the evening to listen to the big national shows. Now those shows were moving from sound to a screen, to be watched instead of heard, and the advertisers were moving with them. Radio did not win that fight. It survived anyway, and it survived by doing something the textbooks on disruption rarely mention: it gave the living room away.

I teach at Singularity University, which Peter Diamandis co-founded, and one vocabulary we use for moments like this is his 6 Ds, a description of how a new technology grows from something nobody notices into something that changes everything. My own framework, the Five Cs of disappearance, describes the same years from the side of the industry being replaced. The 6 Ds explain how a technology grows. The Five Cs explain what it takes away. Read side by side, they show something neither shows alone. An industry facing a new technology has two counts in front of it: the one it reports, and the one that tells it where its customers are going. Storz chose the second. The trouble today is that the second count increasingly belongs to someone else.

The 6 Ds are written from the side of the technology that grows

Diamandis laid out the 6 Ds in Bold, the book he wrote with Steven Kotler in 2015. They describe six stages a technology passes through as it grows, always in the same order, and the clearest way to see them is through the example we use most at Singularity University: reading human DNA.

Figure 1 · animated

Six stages of growth, and what three of them take away

Each stage told through one technology: reading human DNA.

1DigitizedDNADNA becomes four letters: A, C, G, T.What it does becomes information a computer can handle.

2DeceptiveDNASeven years in, about 1% of a genome read.Doubling from tiny numbers looks like nothing.The stretch where the old industry's numbers still look fine.

3DisruptiveDNANew machines leave the old method behind.The doublings add up and an existing market is overtaken.

4DemonetizedDNAOne genome: $95 million in 2001, about $500 in 2022.What used to cost money becomes cheap or free.Compression: someone's price fell to zero.

5DematerializedDNAA sequencer the size of a phone, on USB.The big separate thing shrinks or disappears into something else.Category: someone's product became a feature.

6DemocratizedDNAA saliva test you order online.It ends up in almost everyone's hands.Customer: someone's customers learned to do it themselves.

How to read it. The orange bars show a technology growing the way Diamandis describes it, doubling and doubling again, so the first stages barely register and the last ones fill the frame. Each stage is described from the side of the new technology, with what that stage looked like for DNA in the orange box, and what it means in general below it. The pink lines under the last three read the same stage from the other side of the table, as one of the Five Cs: what the growth takes away from the people who used to sell the thing.

Sources: Peter Diamandis and Steven Kotler, Bold (2015); Diamandis; NHGRI; Bloomberg; MIT Technology Review.

First, a technology is digitized: what it does becomes information a computer can handle. In the 1990s, scientists began turning the chemistry of DNA into long strings of four letters, A, C, G and T, that could be stored and processed like any other data. From that moment, reading DNA could improve at the speed of computing. Next it is deceptive, because growth that keeps doubling starts so small it looks like nothing. The Human Genome Project, which set out in 1990 to read one complete human genome, had by Ray Kurzweil's count read only about one percent of it halfway through, and many skeptics declared it a failure. But one percent that doubles seven times is a hundred percent, and the project finished in 2003, more than two years ahead of schedule.

Then it is disruptive, as new sequencing machines arrived in the 2000s and left the old method behind, and demonetized, as what used to cost a fortune became cheap: by the count the US National Human Genome Research Institute keeps, reading one genome cost about $95 million in 2001 and about $500 by 2022. It is dematerialized when the big separate thing shrinks into something smaller, as when a British company began mailing researchers a sequencer about the size of a phone that runs off a laptop's USB port in 2014, and finally democratized, when anyone can order a DNA test online and post back a tube of saliva.

Figure 2 · animated

Reading one human genome went from $95 million to about $500

How to read it. The orange line is what it cost to read one complete human genome, year by year, as tracked by the US National Human Genome Research Institute. Each step down the scale is ten times cheaper than the one above it. The dashed grey line is where the price would be if it had only fallen as fast as computer chips get cheaper, halving every two years. Until 2007 the two lines travel together. Then new sequencing machines arrive, and by 2022 a genome costs more than a hundred times less than the chips' pace would have made it. That is the demonetized stage, and for the laboratories that sold genome reading at the old price, it is Compression.

Sources: NHGRI, DNA Sequencing Costs: Data, via Our World in Data; MIT Technology Review.

Every one of those words describes the new technology, and that is the right point of view for an entrepreneur or an investor looking for the next one. Read the same words from the other side of the table, though, and each describes something lost. Demonetized means someone's price fell to zero. Dematerialized means someone's product turned into a feature of something people bought for other reasons. Democratized means someone's customers learned to do the work themselves. Those are three of my Five Cs, Compression, Category and Customer, seen from the winner's seat. The 6 Ds are an optimist's map, and a good one. They simply have no column for the people who used to sell the thing.

Diamandis would object, and fairly, that Bold opens with exactly those people. Its first chapter is about Kodak, which built the first digital camera in 1975 and still filed for bankruptcy in 2012, and its lesson for every incumbent is to cannibalize yourself before someone else does. The rest of this essay is partly an answer to that lesson, because the cases I studied suggest it saves something other than what people hope.

The Five Cs are written from the side of the industry being replaced

I first laid out the Five Cs in Nobody Canceled the Iceman, and the study behind this series has sharpened them since, and changed my mind on two of them. Capital is the money. In the Iceman essay I wrote that the money leaves first. The study says something more uncomfortable: new money does arrive early, going to the substitute as customers start to move, but in the industries that disappeared, the old money usually left only after the customers had started to go. Compression is the companies and the margin, as firms merge, close or go bankrupt while the old business shrinks. Those two are signals, the market's reaction, which usually follows once customers are moving. The other three are routes, the paths customers take away. Container, the most common route in the study, is when the need survives but moves to a new form or a new middleman. Customer is when the payment goes around the industry, because the buyer does the work, the supplier sells direct or something free takes its place. Category, the quietest, is when the job is absorbed into something people buy for other reasons. It is the hardest to see, and not, as I once wrote, the one that finishes an industry.

New York's yellow cabs show two of them clearly. The need for a ride survived, and riders moved to a new form: in July 2017, rides booked through apps outnumbered yellow cab trips for the first time. That is Container. Capital moved in both directions. New money went to Uber, which raised $1.2 billion in June 2014, and the old money left through the lenders: LOMTO, a credit union in Queens that had lent to medallion owners, was taken over by regulators in 2017 and liquidated in 2018. Two of five is normal. Most industries in the study that disappeared showed two or three, and none showed all five clearly.

Figure 3 · animated

Five things an industry loses, from the loudest to the quietest

Told through New York taxis, which show two of the five clearly.

SIGNALCapitalThe moneyNEW YORK TAXISUber raises $1.2 billion in 2014. The medallion loses four-fifths of its price, and a lender to its owners is liquidated.New money goes to the substitute, and old money leaves afterward.

SIGNALCompressionThe companies and the marginNEW YORK TAXISWeak in the data: the old firms shrank less clearly than their value did.Firms merge, close or go bankrupt, and margins thin.

ROUTEContainerThe formNEW YORK TAXISRiders stop hailing and start tapping. Apps pass yellow cabs in 2017.The need survives, but customers choose a new form or middleman.

ROUTECustomerThe paymentNEW YORK TAXISNot clear in the data. Most industries show two or three Cs.The buyer does it themselves, the supplier sells direct, or it turns free.

ROUTECategoryThe decisionNEW YORK TAXISNot seen either. Category is the hardest C to see, even looking back.The job is absorbed into something people buy for other reasons.

How to read it. The pink bars show how easy each C is to hear, loudest first. They are not measurements; they show the order. Money and bankruptcies make the news, while the quiet routes are hard to see even afterward. The first two are signals, the market reacting, usually once customers are moving. The last three are routes, the paths customers take away from an industry. No industry needs all five, and none in the study that disappeared showed all five clearly. What matters is whether a route is moving, because no industry disappeared without one.

Sources: TechCrunch; CBS News; American Banker; WNYC.

The deceptive phase looks like a boom from the other side

In May 2011, Uber started taking rides in New York. By Diamandis's definition it was deceptive: a handful of black cars summoned through an app, in a city that ran on yellow cabs. Every yellow cab in New York needs a medallion, the license that lets it pick up passengers who wave from the street, and medallions were bought and sold like property. In my Five Cs, that price is Capital: what it costs to own a piece of the old industry. In April 2013, two years after Uber arrived, the price of a medallion hit an all-time high of $1.3 million. By March 2017 one sold for about $241,000.

Diamandis calls this phase deceptive because the new technology looks too small to matter. Seen from the other side, it is deceptive for a second reason: the old industry's own numbers look better than ever. The taxi owners were not ignoring a threat. Their most important number was telling them they had never been worth more, while their customers were installing an app. The other count was there to read, at the airports and outside the bars late at night, where riders who could have waited in the cab line were already opening an app. That is the cruel part of the deceptive phase. It is when an industry still has time to move, and it is also when everything it measures tells it to stay put.

Figure 4 · animated

The medallion peaked two years after Uber arrived

May 2011Uber arrivesA few black cars, summoned by app

April 2013$1.3 millionThe highest price a medallion ever sold for

March 2017$241,000Four years later

How to read it. A New York medallion is the license that lets a yellow cab pick up passengers who wave from the street, and it is bought and sold like property. Two years after the app arrived, its price set an all-time record. The owners' most important number told them they had never been worth more.

Sources: amNY; CNBC; CBS News.

Radio survived by giving television the living room

Radio never had a quiet early stretch in which to think. By the time Storz bought his station, television was already taking its evenings. National network radio, the coast-to-coast shows paid for by the biggest advertisers in the country, saw its time sales, the airtime it sold to advertisers, fall from $133.7 million in 1948 to about $55 million in 1955, a drop of about 60 percent in seven years. Anyone looking only at that line would have written radio's obituary.

What that line did not show was the rest of the industry quietly changing jobs. Over the same years, local advertising on radio rose from about $171 million to $278 million, and by 1952 local businesses were paying more than half of radio's advertising money. Half of American cars had a radio in 1951, and 80 percent did by 1965. In October 1954 Texas Instruments and Regency put a transistor radio weighing eleven ounces on sale for $49.95. Television needed your eyes and your living room. Radio took the car, the kitchen, the workplace and the pocket, and Storz aimed his stations squarely at those listeners. That is the Container route: the living-room job left for a new form, and radio let it go.

Figure 5 · animated

Radio lost the network and found the town

How to read it. Each line is the airtime radio sold to advertisers, in millions of dollars, in 1948 and in 1955. The orange line is the national networks, the coast-to-coast evening shows that television took. The pink line is local businesses buying time on their town's station. Network radio lost about 60 percent in seven years, while local radio grew by more than half.

Source: Broadcasting, 1956: a quarter century of radio and television time sales.

It would be too neat to call this strategy. Cars and transistors were not radio's inventions, and radio was lucky that cars and workplaces were filling up with listeners so fast. But luck only helps an industry that is willing to give something up. Network radio held on to the evening drama and lost it anyway. The stations that let it go found that people still wanted a voice and a song in the places a screen could not reach, and they were willing to be paid by the shop on the corner instead of the national sponsor.

Most industries that died had a way out too

When I went back through thirty industries, all but two chosen because they faced a substitute,¹ I expected the survivors to be the ones that found a way out and the dead to be the ones that did not. Almost every survivor did have one: a new job, a new payer, something that slowed the substitute down, or the old companies taking over the new form themselves. But so did most of the industries that died. Pagers kept a loyal market in hospitals, where mobile phones lose their signal behind concrete walls. Dial-up internet kept customers in rural homes where broadband never arrived. Both were ways out, and both industries still shrank to a tenth of their peak or less. Of the industries that survived, even as a niche, all but one had found a way out; the exception was home milk delivery, which shrank to a niche without one. So a way out guarantees nothing, and that is the first answer to the question every incumbent asks, what do we do now.

The way out most often recommended is Kodak's lesson: jump into the new technology while you still have the money and the customers to do it. In the industries I studied, the companies that jumped often made it across while the thing they used to sell stayed behind. That is an observation from the cases rather than something the study tested, and jumping did not always save the company either: EarthLink, the dial-up provider, had 1.5 million broadband customers by 2005, but its broadband business never took off against the telephone and cable companies, and in 2016 it was sold to one of them. The yellow directory that once sat beside every telephone shows the pattern most clearly. The publishers moved online early and hard. YP, the largest American publisher, made about a third of its revenue from digital products by 2013 and 44 percent a year later. In Britain, Yell was earning more than four-fifths of its revenue from digital by 2017, and two years later the last printed Yellow Pages in the country was delivered in Brighton. The publishers survived as digital marketing companies, while the book itself disappeared.

The AI wave has produced its own version, and it is more tempting: the old industry sells the new technology what it has. People Inc. (then still called Dotdash Meredith), the publisher of People magazine and dozens of recipe and how-to sites, signed a deal with OpenAI worth at least $16 million a year and later a pay-per-use deal with Microsoft, while Google's share of its traffic fell from 54 percent to 24 percent in two years. The single deals are becoming a standing market. In October 2026, Google's own licensing program passed 400 publications, CNN and the Financial Times among them, in a year when search traffic to news sites fell by 40 percent. In deals like these the new payer is the substitute itself, and what it buys first is the old industry's past. Some of the deals pay every year, but look at what each payment buys. The archive is bought once. What follows is a supplier contract with your replacement, renewed at its price, for as long as it still needs what you make next.

The jukebox now belongs to the substitute

In the early 1950s, the network ledgers told radio that the national sponsors were leaving and were not coming back. They were right, and they were useless for deciding what to do next. Storz found his answer in a jukebox, a machine standing in a room he did not own, taking nickels for songs people had already heard many times. It counted what people kept paying for when nobody was asking them to. The network ledgers were the count radio reported. The jukebox was the count that mattered, and what Storz did was choose which one to believe.

Seventy years later, the car Storz aimed for is being rebuilt by someone else. Broadcast radio still holds 55 percent of the listening Americans do in their cars, a number that still looks like safety. But the people who build the car now own the dashboard. Tesla, BMW and Volkswagen have taken AM out of their electric cars, Tesla's cheapest 2026 models come with no AM or FM at all, and Congress is arguing over a law to force the radio back in. In March, Cumulus, one of the largest radio companies in America, filed for bankruptcy.

The jukebox has moved too, and this is where the radio story stops matching today. Storz could read his jukebox because it stood in restaurants anyone could walk into. Today it plays inside a dashboard radio does not control, on Spotify and the other streaming services that live on that screen. Anyone can see their charts. What only they can see is who played a song again, where they were driving when they did, and what they skipped. The same is true across the AI wave. When People Inc. sells its archive to OpenAI, it sells into the very room where its own jukebox now plays: the questions its readers now ask a chatbot are counted by the buyer, and the count stays there.

So the car has been radio's way out for seventy years, and it was always a lease, not a deed. That is the lesson of the one industry in this essay that survived. Storz did not save radio forever. He read the count in a room he did not own, and then he moved radio to where its listeners were going, while the old money was still leaving. Every industry facing AI has to do the same, and then do it again, because the room it moves into will one day have a new jukebox in it, owned by someone else.

Under every essay in this series is the Iceman Test: eleven questions about your own industry that show which of the Five Cs are already moving, which route your customers are most likely to take, and whether you are still in the stretch where the numbers look fine. Use it to find out which room to watch.

The Five Cs of Disappearance · Self-test

Is your business the iceman?

Industries rarely die because a better rival wins. They disappear because customers quietly move, and in a study of 30 industries, an industry's own numbers were almost always the last place it showed. Eleven questions, about four minutes. Answer for one thing you are paid for, not the whole company.

In a few words, what do customers pay you for?

For example: fix cars, translate documents, sell accounting software, cater weddings.

¹ About the study. The thirty industries are 24 historical cases and 6 software cases. Each was chosen because a substitute reached at least a tenth of its job within ten years, not because of how it ended, so industries that survived and industries that disappeared are both in it; the two exceptions are controls, chosen for heavy consolidation and investors leaving with no apparent substitute, and one of them, freight rail, turned out to have one in trucking. The unit is one job an industry was paid for, such as finding a local business, rather than a company. Each case was scored from 0 to 3 on each of the Five Cs, with a dated source behind every score and the year each one started where it could be dated; a score resting only on qualitative evidence was capped at 2. Its outcome was measured fifteen years after the job's peak: disappeared (down to a tenth of its peak or less), niche, transformed, or held. A way out counts as a new job, a new payer, something that slowed the substitute down, becoming a feature of the substitute, or the old companies capturing the new form. The people who scored the cases never saw the predictions, and the ones who scored the ways out were not told how each industry ended. A second coder independently re-scored eight cases and agreed on the outcome in seven, but could rarely score the quieter routes, Customer and Category, which is a limit on those scores. The limits: most cases were scored by one coder, only one control is clean, the way-out coding was added after the first results, most cases are American, most are media and middlemen, the dates are by year, and the study describes how industries lost their customers, not which of today's will. The full case list and scores will be on the study page.